Crafting Reliable Rules with Robust Control
summary
Applies Robust Control to tune a pricing algorithm for a collateralized lending platform, shares code and design digital twin repos and uses the open source tool cad for simulations, selects PID gains and launches with a P‑only controller on Feb 13 2021, later adds an I term via governance tweaks, monitors outcomes for ~4,000 active users and observes about a $0.30 (~10%) total price drift over three years, with shock tests and worst‑case analyses.
AI-written summary · about
chapters (21)
- 0:00Reliable rules and control
- 1:32Reliable rules in complex systems
- 3:35Control theory for rule design
- 5:17Robust control methodology
- 7:02Rule development process
- 8:03Automated collateralized lending
- 10:34Ideation to implementation in web3
- 11:58Price regulation goals
- 14:05PID controller selection
- 15:10Uncontrollable market factors
- 16:13Structured dynamical models
- 17:39System trajectories from code
- 18:42Robustness testing with shocks
- 19:46Production price controller launch
- 20:50Monitoring system behavior
- 21:52Governance through parameter adjustments
- 23:18Alternative trading strategies
- 24:24Three-year stability results
- 25:26Open-source tools and grants
- 26:28Block Science applications
- 27:53Q&A
related talks
18mKeynote: AI for Financial Services
40mCreating a Contemporary Lending Risk Management System Using Python
55mAlgorithmic Trading and Machine Learning
42mTesting Machine Learning Modules in Production
41mCompetitive Algorithms for Online Control
24mToward Trustworthy AI: Explainability and Robustness
13mScheduled Partial-Credit RL for Reliable Code Generation with Small Language Models (WIP)
19mLessons from Trillion Token Deployments at Fortune 500s